Multinomial Principal Component Logistic Regression on Shape Data

This paper proposes a linear model that uses the principal component scores in shape data and fits the nominal responses in the tangent space of shapes. Multinomial logistic regression for multivariate data and logistic regression for binary responses are considered in this regard. Principal components in the tangent space are employed to improve the estimation of logistic model parameters under multicollinearity and to reduce the dimension of the input data. This paper improves the classification of shape data according to their different nominal groups. Furthermore, we assess the effectiveness of the proposed method using a comprehensive simulation and highlight the benefits of the new method using five real-world data sets.